A NEURAL-NETWORK APPROACH FOR THE DETERMINATION OF INTERHOSPITAL TRANSPORT MODE

We report on the construction of neural networks for determining wheth er pediatric patients requiring transport to a tertiary care center sh ould be moved by air or by ground. The networks were based on the func tional-link net architecture. In two experiments, feedforward supervis ed-learning neural nets were trained with examples of an expert's deci sions and then were used in a consulting mode to provide advice on cas es not previously encountered. Training and validation were performed by a combination of the k-fold cross-validation and leaving-one-out sa mpling methods. Use of the functional-link net rather than the customa ry backpropagation net enabled us to carry out the training with fairl y large amounts of data in realistically short time periods. In the fi rst experiment, capillary refill, skin color, and strider were consist ently the input variables that were most strongly associated with the decision output. In both experiments, the networks were validated by c omparing their performance retrospectively against the determination o f an expert pediatric transport physician. The network was trained bas ed on the expert's opinion about the correct mode of transport for eac h case with error rates of less than 10(-5). (C) 1995 Academic Press, Inc.